Design your first protein with AI
A 90-minute hands-on masterclass for designers who work with living matter, and want to do it with care. It ran live on Thursday 1 October 2026. You can now watch the recording and follow along in your own time.
Why a protein
A fungus coats its spores in a protein that makes water run off them. A silkworm spins a case of protein around itself while its body changes. Some bacteria make proteins that grip the cellulose in plants, so they can break it down and feed.
Each protein starts as a string of letters. The string folds into a shape, and the shape does the job.
Designers meet these proteins later, as mycelium, silk and cellulose. When we make things from living matter, we borrow that work.
AI tools now let you see the shape of a protein in a few minutes, from a browser on your laptop. They make proteins easy to change. They can’t tell you what a change would mean for the organism that makes the protein, for the other organisms it would meet, or for the people who would handle it. You have to ask that yourself. In this session we practised asking, with a real protein in front of us.
What we did
We worked with one protein: a hydrophobin, the water-shedding coat that the fungus Trichoderma reesei wraps around its spores. It is a string of 86 letters.
Everyone folded the real protein in their browser. Then we learned to read the picture: which parts the model is sure about, which parts it is guessing, and what the colours do not tell you.
We built a new version, a double coat: two copies of the protein joined into one chain. Then I took everyone through changes I had run before the session, such as taking out the internal stitches that hold the protein together. Each change got a confidence score, and some of the scores were not what you would expect. A protein with a missing stitch can still score high. Working out why was the main lesson.
Each person picked the version they would take further, and wrote 2 questions. The first was for a scientist: what would they have to test? The second was for themselves: if this protein were ever made, who and what would it touch?
What you get with the recording
The full session, 85 minutes, with chapters, so you can jump to each part.
The slides.
A one-page card with the 5 steps, so you can do it again with a protein from living matter you already work with.
All the sequences we used, plus ones I picked for algae, mycelium and textile projects.
A list of the free tools and databases, with what each one is for.
After you pay, you go straight to a page with the video and the files on it. Bookmark it, so you can come back any time.
What you will be able to do
Fold a protein from its letters, in your browser, for free.
Read the confidence colours and the score.
Change a protein on purpose and compare the versions.
Open the shape in a free viewer and export a file for 3D printing.
You’ll also know what the picture doesn’t tell you. A protein that would fall apart the moment it met water can get the same confident picture as one that would hold.
Confidence is the model’s view of its own drawing. It doesn’t test anything in the world, and it knows nothing about the organism that would have to make it.
The tools, and what you need
Get the recording
Library members: the recording is included in your membership. Look for it in your member email.
The recording does not include feedback on your own result. That was only part of the live session.
Teaching a course or a studio, and want to run this exercise with your students? Write to [email protected].
Want to go further?
From 5 November to 10 December 2026. It starts from zero: what a protein is, where our knowledge of proteins lives, how the AI tools work, and what the newer generative tools can do. Every week you apply it to a project of your own, and you finish with a protein design proposal and a certificate.
Three live sessions, on Thursdays at 17:00 CET. Founding prices until Sunday 25 October: Student $195, Standard $350.
Questions people ask
No. It ran on Thursday 1 October 2026. The recording has the whole session, and you can follow every step in your own time.
No. Most of the session is spent folding and changing a real protein and noticing what happened. The questions about care come out of that work.
No. You can watch it through first, and do the steps afterwards with the card.
No. We change a protein that already exists in nature. Designing new proteins from scratch is part of the six-week course.
No. If you can paste text into a box and press a button, you can do every step. I explain each biology word the first time I use it.
Yes. The AI model runs on the tool’s servers, not on your machine. You need a browser and a working connection.
On 1 October 2026, a free Rowan account had enough credits to follow every step. Tools change their free plans, so check when you sign up.
The page is yours to come back to. Bookmark it after you pay.
No. You already have it. Check the email I sent you after the session.
AI-Native Biodesign: Proteins, a six-week course from 5 November. See the course.
Who’s teaching
I’m Raphael, and I started in the lab. During my biotechnology degree at UCL I spent a year at Roche in Basel, folding and purifying proteins, and I wrote my final-year thesis on protein folding and drug design. Later I was the lab manager of a malaria research lab at Imperial College London.
Then I moved into design: an MA at the Royal College of Art, a PhD at Queen Mary University of London, and a postdoc on living matter at TU Delft. I’ve taught design students at the RCA as a visiting lecturer.
Now I work where the two meet. In 2026 I won compute time on Isambard-AI, part of the UK’s national AI research resource, and used it to test how AI protein predictions hold up across about 1,700 structures. In 2025 I ran AlphaFold Futures, a workshop for designers from 5 universities. In January 2026 I led a week-long international design workshop at Shih Chien University in Taipei. And I run Biodesign Academy, where I write about how living matter works, from the molecule up.
Questions? Reply to any Biodesign Academy email, or write to [email protected].
